Resolving Phenotyping Discordance with SPACEMAP Machine Learning Framework

Introduction and Purpose: Multiplex imaging provides powerful insight into cellular organization, yet the complexity of these datasets requires robust analytical tools to extract meaningful biological information. The purpose of this study was to develop a unified analytical framework enabling reliable, high-resolution characterization of the tissue microenvironments in multiplex images. To achieve this, we developed SPACEMAP (Spatial Phenotyping And Classification with Enhanced Multiplex Analysis Pipeline), a comprehensive Python and Qupath-based platform that integrates image registration, cell segmentation, quality check, artifact removal, tissue and zone classification, spatial feature extraction, and a consolidated phenotyping approach into a single workflow. Methods: To determine an optimal cell-classification strategy, we benchmarked our phenotyping method, RESOLVE, against three established approaches, Leiden clustering, Self-Organizing Maps, and SCIMAP. This evaluation revealed substantial inconsistencies among existing methods. To address this, SPACEMAP incorporates two complementary workflows: (1) a machine learning model trained on expert-annotated cells, and (2) a consensus-based model that integrates high-confidence cell assignments across methods, enabling robust classification even when manual references are limited. Summary: We applied SPACEMAP to newly generated multiplex imaging datasets from colorectal tissue samples and further evaluated performance using a publicly available dataset. These analyses demonstrate that SPACEMAP improves classification consistency, reduces variability introduced by method selection, and supports reproducible extraction of spatial features for further downstream analysis. Conclusions: SPACEMAP provides a standardized, high-fidelity workflow for spatial phenotyping that minimizes reliance on labor-intensive manual annotation and improves reproducibility in multiplex imaging studies. Its design supports adaptation to evolving imaging technologies and marker panels, enabling researchers to more effectively investigate tissue organization and generate biologically meaningful insights. Citation Format: Arely Perez Rodriguez, Bassel Dawod, Sebastian Diegeler, Eslam A. Elghonaimy, Megan Wachsmann, Purva Gopal, David Hein, Paul H. Acosta, Andrew Jamieson, Gaudenz Danuser, Robert Timmerman, Satwik Rajaram, Todd A. Aguilera, . Resolving phenotyping discordance with SPACEMAP, an integrated machine learning framework [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 5498.